Windows: portable intrinsics

[h-vetinari]

Trying to compile windows for AVX2 in conda-forge/faiss-split-feedstock#27
(after #1600) surfaced a bunch of things (#1680, #1681, #1682), but the most voluminous problem
was MSVC being much worse at dealing with operator overloads and casts around __m128 / __m256.

This lead to loads of errors that looked as follows:

[...]\faiss\utils\distances_simd.cpp(411): error C2676: binary '+=': '__m128' does not define this operator or a conversion to a type acceptable to the predefined operator
[...]\faiss\utils\distances_simd.cpp(440): error C2676: binary '-': '__m256' does not define this operator or a conversion to a type acceptable to the predefined operator
[...]\faiss\utils\distances_simd.cpp(441): error C2676: binary '*': 'const __m256' does not define this operator or a conversion to a type acceptable to the predefined operator
[...]\faiss\utils\distances_simd.cpp(446): error C2676: binary '+=': '__m128' does not define this operator or a conversion to a type acceptable to the predefined operator
[...]\faiss\utils\distances_simd.cpp(451): error C2676: binary '-': '__m128' does not define this operator or a conversion to a type acceptable to the predefined operator
[...]\faiss\utils\distances_simd.cpp(452): error C2676: binary '*': 'const __m128' does not define this operator or a conversion to a type acceptable to the predefined operator
[...]\faiss\utils\distances_simd.cpp(459): error C2676: binary '-': '__m128' does not define this operator or a conversion to a type acceptable to the predefined operator
[...]\faiss\utils\distances_simd.cpp(460): error C2676: binary '*': '__m128' does not define this operator or a conversion to a type acceptable to the predefined operator
[...]\faiss\utils\distances_simd.cpp(471): error C2440: '<function-style-cast>': cannot convert from '__m256i' to '__m256'

I've followed https://software.intel.com/sites/landingpage/IntrinsicsGuide/ to try to replace everything correctly,
but this will surely require close review, because I'm not sure how well these code-paths are checked by the
test suite.

In any case, with the commits from #1600 #1666 #1680 #1681 #1682, I was able to build libfaiss & faiss
for AVX2 on windows (while remaining "green" on linux/osx, both with & without AVX2).

Sidenote: the issues in the last commit (26fc7cf)
were uncovered by adding the __SSE3__ compat macros in #1681.

[h-vetinari]

The failing test seems to be flaky.

[beauby]

Looking good, @mdouze could you take a look?

[mdouze]

I will take a look. The contbuild system on Linux does not compile with AVX2 enabled...

[mdouze]

Looks good to me.

[facebook-github-bot]

@mdouze has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

[h-vetinari]

@mdouze, would you prefer me to rebase this on master after #1690? Or anything else I can help with? I'm kinda waiting with conda-forge/faiss-split-feedstock#27 until this goes in because I don't want to get this patch wrong.

[facebook-github-bot]

@mdouze merged this pull request in 1afaddb.